text
stringlengths
1
2.28M
arxiv_id
stringlengths
9
12
url
stringlengths
31
34
timestamp
stringdate
2007-02-27 04:31:34
2023-03-06 02:13:47
yymm
stringdate
0701-01-01 00:00:00
2302-01-01 00:00:00
language
stringclasses
8 values
Qwen3_token_count
int64
1
2.08M
uuid
stringlengths
36
36
Llama3.2_token_count
int64
1
1.03M
\section{Introduction} The idea of constructing a confidence set of statistically plausible models and picking a policy that maximizes the expected return in the best of these models can be traced back to the pioneering work of \citet{LR85} in the context of multi-armed bandit problems, and has been successfully exten...
2302.14004
https://arxiv.org/abs/2302.14004
2023-03-06 02:13:47
2302
en
28,469
c3159a44-af95-4982-99e3-9b25d3c0d439
28,369
\section{Independent linear function approximation} \label{sec:linear-assumption} \subsection{Assumption for Independent Linear Function Approximation} We want an assumption that meets the three following criteria: \begin{itemize} \item Does not trivialize the problem (imply essentially tabular or essentially sing...
2302.06606
https://arxiv.org/abs/2302.06606
2023-03-03 02:08:54
2302
en
70,899
7444706d-c93b-4923-aceb-6e8b55ee708e
70,587
\section{Introduction} Decision-making under uncertainty in a multi-agent system has shown its potential to approach artificial intelligence, with superhuman performance in Go games \citep{silver2017mastering}, Poker \citep{brown2019superhuman}, and real-time strategy games \citep{vinyals2019grandmaster}, etc. All the...
2302.03673
https://arxiv.org/abs/2302.03673
2023-03-03 02:08:54
2302
en
23,624
153c58b0-18fe-403f-a87c-fc13ab741f1a
23,266
\section{Introduction} The Large Hadron Collider (LHC) \citep{Evans:2008zzb}, the most energetic particle collider in human history, collides highly-energetic protons to examine the underlying physics of subatomic particles and extend our current understanding of the fundamental forces of nature, summarized by the Sta...
2302.00695
https://arxiv.org/abs/2302.00695
2023-03-02 02:22:42
2302
en
10,140
f7351840-1582-458d-85e2-58040f26fe26
9,797
\section{Introduction} Multivariate non-parametric regression can be approached from a variety of methods: decision trees, local averaging methods such as Nadaraya-Watson estimation or $k$-nearest-neighbor regression, neural networks, and methods based on positive definite kernels such as smoothing splines, kriging, a...
2302.03459
https://arxiv.org/abs/2302.03459
2023-03-02 02:18:17
2302
en
14,053
87a6a8be-6b46-4129-bb3a-c056bc945872
13,937
\subsubsection*{\bibname}} \begin{document} \twocolumn[ \aistatstitle{Statistical Analysis of Karcher Means for Random Restricted PSD Matrices} \aistatsauthor{ Hengchao Chen \And Xiang Li \And Qiang Sun } \aistatsaddress{ Department of Statistical Sciences\\ University of Toronto \And School of Mathematical S...
2302.12426
https://arxiv.org/abs/2302.12426
2023-03-02 02:03:00
2302
en
14,623
65f4fa50-d689-47d6-92c9-cf79eba93ea5
14,475
\section{Introduction} Learning halfspaces in the presence of noise is one of the most basic and well-studied problems in computational learning theory. A large body of work has obtained results for this problem under a variety of different noise models and distributional assumptions (see e.g.\ \cite{balcan2021noise}...
2302.14853
https://arxiv.org/abs/2302.14853
2023-03-01 02:24:18
2302
en
23,566
6b07be20-c12a-40b2-96e0-45017d4457e6
23,433
\section{Introduction} \label{sec:intro} \input{section/introduction.tex} \subsection{Preliminaries} \input{section/prelims.tex} \subsection{Our results} \input{section/results.tex} \section{Technical overview} \label{sec:tech} \input{section/techniques.tex} \section{Learning with conditional probabilities (Theorem ...
2302.14753
https://arxiv.org/abs/2302.14753
2023-03-01 02:21:47
2302
en
44,084
c455751c-557b-4984-ba26-6ae180bd4c4f
44,004
\section{Introduction} Recommendation systems are an essential component of modern online platforms. They help individuals find candidates to interview for job postings, dating partners, products to try, or media to engage with. Even though most users have recurring relationships with recommendation systems, nearly a...
2302.03561
https://arxiv.org/abs/2302.03561
2023-03-01 02:21:36
2302
en
45,207
c778f5bb-5b5b-43d2-9bd4-58736ad64885
44,942
\section{Introduction} Machine learning methods -- often consisting of \emph{artificial neural networks (ANNs)} trained through \emph{gradient descent} (GD) type optimization methods -- are nowadays omnipresent computational methods which are heavily employed in many industrial applications as well as scientific r...
2302.14690
https://arxiv.org/abs/2302.14690
2023-03-01 02:20:50
2302
en
38,588
246d2700-8b10-400a-87ec-b3f9ed5cb478
38,319
\section{Deferred proofs of Section~\ref{sec:guar}} \label{sec:app:guar} We first present the proof of Lemma~\ref{lem:algo:bound}. \begin{proof}[Proof of Lemma~\ref{lem:algo:bound}] We start the proof by reminding the two no-regret guarantees of Algorithm~\ref{algo:1}. \eqref{eq:prel:Rmax} and \eqref{eq:prel:Rmin...
2302.14686
https://arxiv.org/abs/2302.14686
2023-03-01 02:20:44
2302
en
23,326
7c3027e9-6352-44a4-bb2e-60a6f19026fe
23,274
\section{Introduction} \label{sect:intro} In the past decade, the training and inference costs of modern deep neural networks (DNNs) are gradually becoming prohibitive \citep{resnet, vit, liu2021swin}, especially for large language models \citep{GTP3}. Among all these large models, generative adversarial networks (GAN...
2302.14670
https://arxiv.org/abs/2302.14670
2023-03-01 02:20:07
2302
en
9,243
b6e383f1-5c05-4592-ae1a-bc3b4c735ae1
8,954
\section{Introduction} Machine learning is an increasingly dominant part of software systems and, thus, is of increasing interest in the software testing community. Compared to the standard statistical performance analysis on independent, identically-distributed data that machine learning practitioners and researchers...
2302.14616
https://arxiv.org/abs/2302.14616
2023-03-01 02:18:50
2302
en
7,251
7f02f285-0b19-432a-8caa-b9f98230ce26
7,048
\section{Introduction} \label{sec:intro} Clustering is an important unsupervised learning technique with applications in many areas including information retrieval \citep{jardine1971use}, image segmentation \citep{coleman1979image}, pattern recognition \citep{diday1981clustering}, data mining \citep{mirkin2005clusterin...
2302.14599
https://arxiv.org/abs/2302.14599
2023-03-01 02:18:29
2302
en
12,718
cad5bea3-04a3-4d96-be93-6952c0451a1e
12,280
\section{Introduction} \label{sec1} Time series of flows and velocities in traffic systems exhibit fluctuations that indicate and reflect non-stationarity. Nevertheless, traffic systems organize themselves in different operational modes that exist for certain time spans. Put differently, there are quasi-stationary stat...
2302.14596
https://arxiv.org/abs/2302.14596
2023-03-01 02:18:27
2302
en
9,604
fca81465-2b6e-4c52-8b6c-01788975e0c2
9,304
\section{Introduction} Data preprocessing, which involves data labeling, is often the most expensive aspect of deploying machine learning\cite{polyzotis2018data}. Active learning is a sub-field of machine learning that proposes a solution to this problem; it is concerned with selecting which unlabeled datapoints would...
2302.14567
https://arxiv.org/abs/2302.14567
2023-03-01 02:17:27
2302
en
11,496
8bddffef-e8a3-45ab-948b-960523d37d8c
10,734
\section{Introduction} Recent advances in deep neural networks have demonstrated remarkable performance in a wide variety of applications, ranging from recommendation systems and improving user experience to natural language processing and speech recognition \cite{Abiodun2018StateoftheartIA}. Nevertheless, blindly rely...
2302.14552
https://arxiv.org/abs/2302.14552
2023-03-01 02:17:00
2302
en
15,188
ea1da8bb-4090-45a6-a699-9f5272ab4093
14,261
\subsection{Bayesian Adaptive Design} \label{sec:BAD} BED can be extended to adaptive settings---which we will refer to as \emph{Bayesian adaptive design} (BAD)---by iteratively incorporating information gathered during the experiment. In particular, if our designs and outcomes can be broken down into a sequence of st...
2302.14545
https://arxiv.org/abs/2302.14545
2023-03-01 02:16:52
2302
en
8,932
e7629d10-cd0c-41a5-a2ee-33e283c3bd68
8,521
\section{\@startsection {section}{1}{\z@}{3.25ex plus 1ex minus .2ex}{1.5ex plus .2ex}{\large\bf}} \def\subsection{\@startsection{subsection}{2}{\z@}{3.25ex plus 1ex minus .2ex}{1.5ex plus .2ex}{\normalsize\bf}} \@addtoreset{equation}{section} \makeatother \title{Maximum Likelihood With a Time Varying Parameter} ...
2302.14529
https://arxiv.org/abs/2302.14529
2023-03-01 02:16:22
2302
en
8,875
ea667a53-5a42-4284-9c09-101222c0a43c
8,840
\section{Introduction} One of the key challenges in machine learning research concerns the ``generalization'' behavior of learning algorithms. That is: if a learning algorithm performs well on the training set, what guarantees can one provide on its performance on new samples? While the question of generalization is ...
2302.14518
https://arxiv.org/abs/2302.14518
2023-03-01 02:16:03
2302
en
12,757
5bffd7fb-1557-446b-90e2-952c2263497d
12,740
\section{Omitted Proofs and Derivations} \label{app:proofs} \subsection{\cref{sec:background}} \label{app:background} First, we provide an explanation on the range of disagreement without normalization: \begin{proposition}[Range of non-normalized disagreement] The expression $\Pr[f_{\theta}(x) \neq f_{\theta'}(x)]$ h...
2302.14517
https://arxiv.org/abs/2302.14517
2023-03-01 02:16:03
2302
en
16,039
2ae95e76-51f1-48bd-8943-9a3ce7f0bad0
15,765
\section{Introduction} For many applications in sciences and engineering, such as emulation-based studies, design of experiments, and automated machine learning, the goal is to optimize a complex black-box function $f(\boldsymbol{x})$ in a $D$-dimensional space, for which we have limited prior knowledge. The main ch...
2302.14510
https://arxiv.org/abs/2302.14510
2023-03-01 02:15:43
2302
en
14,773
41146958-4765-416e-b0b2-8529f596a21b
14,389
\section{Effect of pre-training method} \label{appx:pretrain} Table~\ref{tab:pretrain} presents the test accuracy of PAWS and RoPAWS, initialized from ImageNet supervised or SwAV~\citep{caron2020unsupervised} pre-trained models. We used SwAV for our main table since it gives a better initialization. Nevertheless, RoP...
2302.14483
https://arxiv.org/abs/2302.14483
2023-03-01 02:14:58
2302
en
26,445
c712c303-4e2e-433e-b1bd-836aac657138
25,734
\section{Introduction} This paper focuses on the resolution of the so-called ``\(\ell_0\)-regularized least-squares problem'' given by \begin{align} \stepcounter{equation} \tag{\theequation-\(\mathcal{P}\)} \label{prob:prob-base} \min_{\mathbf{\pvi} \in \kR^{n}} \ \mathrm{P}(\mathbf{\pvi}) \triangleq \tfrac...
2302.14471
https://arxiv.org/abs/2302.14471
2023-03-01 02:14:17
2302
en
18,893
2bb590a8-b5ce-42a2-965a-771f184f9157
18,776
\section{Introduction} Scientific machine learning (Sci-ML) \cite{E_CICP,Edeep18,raissi2019physics,sirignano2018dgm,zang2020weak} has gained popularity in recent years for solving partial differential equations (PDE). The Sci-ML methods are different from conventional ML methods in that they can take into account phys...
2302.01529
https://arxiv.org/abs/2302.01529
2023-03-01 02:14:10
2302
en
15,938
92784afa-3c76-4368-a325-8cc669accc5a
15,649
\section{Introduction} Interacting particle systems and related mean field models have been recently used in a large variety of challenging domains, including biology, social systems and economics, see e.g. \cite{bellomo2017active,gibelli2019crowd,gibelli2020crowd2}. These models originated in statistical mechanics, ...
2302.14446
https://arxiv.org/abs/2302.14446
2023-03-01 02:13:37
2302
en
30,411
e7c6d904-cb0e-4d9c-89b8-90f0ab415354
29,345
\section{Introduction}\label{SecIntro} Covariate shift adaptation has been a pivotal part of transfer learning.\cite{2007Covariate} It is a prevalent setting for machine learning in which the feature distributions differ between the train (source) and test (target) domains, while the conditional distributions of the o...
2302.14427
https://arxiv.org/abs/2302.14427
2023-03-01 02:13:02
2302
en
15,008
335fb9e3-aa6f-45ff-9fd3-af45c90ad5c9
14,379
\section{Introduction} In recent years, due to the success of Reinforcement learning (RL) techniques, there has been a renewed interest in the study of Markov decision processes (MDP) which is the underlying mathematical framework for RL. A number of applications of RL are safety-critical (such as robotics \cite{kobe...
2302.13152
https://arxiv.org/abs/2302.13152
2023-03-01 02:12:26
2302
en
21,512
7333052f-bf29-468d-b88e-5e9e49b9bb34
21,254
\section{Introduction}\label{sec: intro} \input{text/1.Introduction.tex} \section{Preliminaries}\label{sec: prelim} \input{text/2.Prelim.tex} \section{Thompson sampling and the choice of priors}\label{sec: TS} \input{text/3.TS.tex} \section{Main results}\label{sec: main} \input{text/4.rslt.tex} \section{Experiments...
2302.14407
https://arxiv.org/abs/2302.14407
2023-03-01 02:12:24
2302
en
44,646
0fa2e6d0-a0fe-486d-b893-72784876cf9c
44,460
\section{Introduction} Arguably, Deep Neural Networks (DNN) are one of the main drivers of recent advancements in machine learning, leading to state-of-the-art performance in domains such as image classification~\citep{resnets,Tan2019EfficientNetRM} and natural language processing~\citep{devlin2018bert,liu2019roberta}....
2302.11836
https://arxiv.org/abs/2302.11836
2023-03-01 02:11:07
2302
en
15,826
9c9fc628-f165-4ab8-83f5-92146bcc6730
15,676
\section{Introduction} Sparse graph recovery is an important tool to gain insights from data and is a widely researched topic~\cite{heckerman1995learning,bnlearn,shrivastava2022methods}. Given an input data $X$ with $M$ samples and $D$ features, the graph recovery algorithms discover the feature dependencies in the f...
2302.13582
https://arxiv.org/abs/2302.13582
2023-03-01 02:09:52
2302
en
12,909
62ccb3a5-b304-4adf-862d-f7f7df310e3b
12,436
\section{Introduction} Large-scale dynamic systems arise in many scientific fields like engineering, biomedicine and neuroscience, examples being sensor networks, power grids and the brain \cite{Hagmann2008,Hickman2017,Pagani2013,Prando2020}. Modeling these complex physical systems, starting from input-output data, is ...
2302.10959
https://arxiv.org/abs/2302.10959
2023-03-01 02:09:48
2302
en
13,986
88d4ec51-5607-438b-bf2d-6bbfed790433
13,739
\section{Introduction} Generative adversarial networks (GANs) have become a crucial data-driven tool for generating synthetic data. GANs are generative models trained to produce samples from an unknown (real) distribution using a finite number of training data samples. They consist of two modules, a generator G and a d...
2302.14320
https://arxiv.org/abs/2302.14320
2023-03-01 02:09:28
2302
en
10,791
4cd47a29-b629-4ad3-80b3-4eed55884fc5
10,643
\section{Introduction} \label{sec:intro} Langevin dynamics frequently appear in sampling methods for high dimensional, non-Gaussian probability distributions. Given access to the gradient of the logarithm of the target density, many Markov chain Monte Carlo (MCMC) methods use Langevin dynamics as an core ingredient. ...
2302.07227
https://arxiv.org/abs/2302.07227
2023-03-01 02:07:18
2302
en
22,847
f29eb0d2-c8e5-415b-a694-6d714c952603
22,418
\section*{Acknowledgements} The authors thank Ian Waudby-Smith for insightful discussion and review. \bibliographystyle{plainnat} \section{Confidence Sequences for Fixed $v$} \label{app:confseqreview} Since our algorithm operates via reduction to pointwise confidence sequences, we provide a brief self-contained rev...
2302.14248
https://arxiv.org/abs/2302.14248
2023-03-01 02:06:52
2302
en
25,957
64e4a3ca-4474-4ee5-a32e-55b0118f26c9
25,722
\section{Introduction} \subsection{Context} This paper is concerned with the problem of sampling a probability distribution (the target) known up to normalization. This problem is fundamental in many applications arising in computational science and engineering and is widely studied in the applied mathematics, machi...
2302.11024
https://arxiv.org/abs/2302.11024
2023-03-01 02:04:55
2302
en
44,762
b2994ceb-5fb3-4566-892f-c827a9cbcc65
44,295
\section{Introduction} In problems with limited context-specific labeled data, machine learning models often fail to generalize well. Modern machine learning approaches such as transfer learning~\citep{pan2010survey, weiss2016survey}, domain adaptation~\citep{ sun2015survey}, meta-learning~\citep{vilalta2002persp...
2302.14186
https://arxiv.org/abs/2302.14186
2023-03-01 02:04:15
2302
en
19,585
6329c264-454c-448e-85b7-6146778c2c2b
19,243
\section{Proofs for~\cref{sec:lower-bounds}} \subsection{Proof of~\cref{thm:lb-adaptive-adv}} \label{sec:apdx-thm-lb-adaptive-adv} We build on the following property of the padded Tardos code as done in finger-printing lower bounds. Given a matrix $X \in \{-1,+1\}^{(n+1) \times p}$, we say that $j \in [p]$ is a cons...
2302.14154
https://arxiv.org/abs/2302.14154
2023-03-01 02:02:42
2302
en
57,996
bf5db5ad-70aa-4d16-93ba-97071887f8d0
57,691
\subsection{Derivation of the 1-RSB free entropy}\label{subsec_app:1rsb_derivation} We perform here, for completeness of our presentation, the textbook calculation of the spherical perceptron free entropy at the one-RSB level. We start again from eq.~\eqref{eq:Phir_final}, which we rewrite using a Gaussian transformat...
2302.14112
https://arxiv.org/abs/2302.14112
2023-03-01 02:01:25
2302
en
66,867
f02353a8-2474-414a-9ff7-74ed186c349b
66,441
\section{Introduction}\label{sec:intro} The field of Artificial Intelligence (AI) has experienced remarkable growth over the past decade, particularly in Natural Language Processing (NLP). Current state-of-the-art NLP applications, such as translation or text-based recommendations, rely heavily on Deep Neural Network...
2302.14063
https://arxiv.org/abs/2302.14063
2023-03-01 02:00:10
2302
en
26,812
1a8bafd3-2d0e-4478-9d28-7cd0921919f2
26,390
\subsubsection*{\bibname}} \definecolor{shadecolor}{gray}{0.9} \declaretheoremstyle[ headfont=\normalfont\bfseries, notefont=\mdseries, notebraces={(}{)}, bodyfont=\normalfont, postheadspace=0.5em, spaceabove=1pt, mdframed={ skipabove=8pt, skipbelow=8pt, hidealllines=true, backgroundcolor={shadecolor}, inner...
2302.14043
https://arxiv.org/abs/2302.14043
2023-02-28 02:36:36
2302
en
18,798
c59272b9-a188-48a5-972e-61688973d407
18,519
\section{Proofs}\label{sec:appendix_proofs} \section{Experiments} \label{sec:appendix_experiments} We implement all experiments in Pyro \citep{pyro}, which is a probabilistic programming framework on top of PyTorch \cite{paszke2019pytorch}. Our code will be open-sourced upon publication. \subsection{Computing eval...
2302.14015
https://arxiv.org/abs/2302.14015
2023-02-28 02:35:54
2302
en
24,872
41085dd4-623f-49bb-ab76-b7cb17a0077d
23,812
\subsubsection*{\bibname}} \bibliographystyle{abbrvnat} \usepackage{calc \usepackage{subcaption} \usepackage{authblk} \theoremstyle{plain} \newtheorem{theorem}{Theorem} \newtheorem{corollary}{Corollary} \newtheorem{definition}{Definition} \newtheorem{condition}{Condition} \newtheorem{lemma}[theorem]{Lemma} \Declar...
2302.14011
https://arxiv.org/abs/2302.14011
2023-02-28 02:35:49
2302
en
12,500
1a430a75-c09f-4931-b7bc-6513f8bfa6ac
12,307
\section*{Background} Matrix-assisted laser desorption/ionization imaging mass spectrometry (MALDI IMS), also known as MALDI Imaging, is a label-free tool for spatially furnishing molecular weight information of compounds like proteins, peptides, and many others (see, e.g., \cite{Mortier2021}, \cite{Caprioli}). Provid...
2302.13948
https://arxiv.org/abs/2302.13948
2023-02-28 02:33:53
2302
en
14,250
29f13fb6-0fd1-444e-b7c5-11bc59e21765
13,902
\section{Experiment Details} \label{app:exp} This section describes various experiment details regarding model architectures and training hyperparameters. \subsection{Regression} \label{sub_app:regression} \begin{figure}[t] \centering \includegraphics[width=0.3\linewidth]{figures/regression2.png} \includeg...
2302.13934
https://arxiv.org/abs/2302.13934
2023-02-28 02:33:41
2302
en
122,070
ca6e47a4-b6d9-49d6-8157-a1be280ac6d5
121,682
\subsubsection*{\bibname}} \begin{document} \runningtitle{Efficient Informed Proposals for Discrete Distributions via Newton’s Series Approximation} \runningauthor{Yue Xiang, Dongyao Zhu, Bowen Lei, Dongkuan Xu, Ruqi Zhang} \twocolumn[ \aistatstitle{Efficient Informed Proposals for Discrete Distributions \\via Ne...
2302.13929
https://arxiv.org/abs/2302.13929
2023-02-28 02:33:33
2302
en
12,098
f9281132-5679-4692-b921-8a2c10fec038
11,879
\section{{Additional proofs regarding numerical approximations.}} \section{Dataset description}\label{appendix:Dataset Description} We provide a brief description of the datasets and models used for the experiments. The models used for Bayesian logistic regressions were taken from the UCI Machine Learning Repository ...
2302.13918
https://arxiv.org/abs/2302.13918
2023-02-28 02:33:12
2302
en
35,088
88e34ad9-b55c-4455-8af3-2d4bbdd9eccd
32,603
\section{Introduction} \label{sec:introduction} Recently, much effort has been put into creating decision-making systems based on machine learning for various high-risk applications, such as financial operations, medical diagnostics, autonomous driving, \emph{etc.} These systems should comprise several important prope...
2302.13875
https://arxiv.org/abs/2302.13875
2023-02-28 02:32:15
2302
en
10,700
02851922-28c4-44c8-9938-40f41f06615f
10,556
\section{Comparison to Related Work} Please refer to \autoref{tab:compare-dokhorn} for a summary of details that differentiate our work from that of \citet{dockhorn2022differentially}. \begin{table}[h] \setlength\tabcolsep{4pt} \begin{center} \begin{small} \begin{tabular}{lccc} \toprule & \ci...
2302.13861
https://arxiv.org/abs/2302.13861
2023-02-28 02:31:42
2302
en
13,412
02d6d403-4621-4986-8c7f-acf791dcd480
12,756
\section{Introduction} Let $\pi$ be a probability density on $\mathbb{R}^d$ of the form \begin{equation} \pi(x)=\frac{\gamma(x)}{Z},\qquad Z=\int_{\mathbb{R}^d} \gamma(x) \mathrm{d}x, \end{equation} where $\gamma:\mathbb{R}^d \rightarrow \mathbb{R}^{+}$ can be evaluated pointwise but the normalizing constant $Z$ is...
2302.13834
https://arxiv.org/abs/2302.13834
2023-02-28 02:30:54
2302
en
13,978
218aa0e0-eb1c-4d5a-8e11-201ea37bd072
13,742
\section{Introduction} \label{sec:new} Geospatial applications in ecology, climatology, agriculture, forestry, environmental health often involve data that is either naturally binary, e.g., abundance of species \citep{spnngppaper} or incidence of diseases in precision agriculture \citep{zhang2002}, or is dichotomized b...
2302.13828
https://arxiv.org/abs/2302.13828
2023-02-28 02:30:52
2302
en
22,264
276171c9-96dc-45ca-acc0-27bd2579e2e5
22,034
\section{Introduction} Non-equilibrium fluctuations in quantum systems arise when a system is driven away from thermal equilibrium, either by applying an external field or through interactions with other systems. In such situations, the system's energy and other properties fluctuate randomly over time leading to a dis...
2302.13717
https://arxiv.org/abs/2302.13717
2023-02-28 02:28:16
2302
en
12,385
9582e343-77c2-4f10-8a2b-d4b1afeabdcc
11,976
\section{Algorithms and Results} \label{sec:algorithm} The pseudocode of our algorithm is presented in \pref{alg:main_alg}. It is based on the general $\textsc{FTRL}\xspace$ framework, which finds $p^t$, the distribution of selecting arms in around $t$, via solving the optimization problem $p^t = \argmin_{p \in \Omeg...
2302.13534
https://arxiv.org/abs/2302.13534
2023-02-28 02:22:57
2302
en
104,627
183528df-e7a7-4ee2-aedb-5cd3c8d32c16
104,353
\section{Introduction} Bagging and its variants are some of the most commonly used randomized ensemble methods in statistics and machine learning. These methods combine multiple models fitted on different bootstrapped or subsampled datasets to produce a more accurate and stable prediction \citep{breiman_1996,pugh_2002...
2302.13511
https://arxiv.org/abs/2302.13511
2023-02-28 02:22:12
2302
en
20,482
148da664-7391-4b44-900f-32cd4f421cb4
20,215
\section{Introduction}\label{sec:intro} \IEEEPARstart{D}{eep} learning has achieved excellent results in a variety of different optimization tasks, such as the inference problems \cite{alexnet,maas,resnet} and the generative modeling problems \cite{vae,gan,radford}. Although the traditional domain of deep learning h...
2302.12304
https://arxiv.org/abs/2302.12304
2023-02-28 02:21:06
2302
en
16,062
ddd4df5f-47bb-4c82-ad25-f42cd6f2cbe9
15,930
\section{Application domains} In this section, we discuss several application domains of uncertainty quantification for deep learning models. For each application domain, we discuss the motivation for developing uncertainty-aware models and the nature of the data, the source of uncertainty, and the challenges associ...
2302.13425
https://arxiv.org/abs/2302.13425
2023-02-28 02:19:08
2302
en
30,368
b07fb929-b271-444e-9f68-8620973fb56b
29,851
\section{Introduction} \IEEEPARstart{50}{years,} ago, Francis Anscombe published a groundbreaking paper ``Graphs in Statistical Analysis'', in which he argued that data visualisation is a crucial element for statistical analysis (see \cite{Anscombe73}). He showed four sets of data, later called ``Anscombe's quartet'' ...
2302.13356
https://arxiv.org/abs/2302.13356
2023-02-28 02:17:20
2302
en
1,220
e4c9ca19-9f20-41f0-b722-644459ad1933
1,209
\section{INTRODUCTION}\label{chap:intro} The offline contextual bandit is a simple but powerful model for decision-making with a wide range of applications such as data-driven personalized medical treatment, recommendations, and advertisements on online platforms. In the evaluation of its policy value, the inverse...
2302.13348
https://arxiv.org/abs/2302.13348
2023-02-28 02:17:12
2302
en
37,628
05b85f6e-910d-49e8-97c7-6f5c283e2f2f
37,405
\section*{APPENDIX} \section{ADDENDUM TO SECTION~\ref{sec:methods}} \subsection{Problem~\eqref{eq:fair_PCA} May Not Be Well Defined}\label{app:not_well_defined} Let $n=2n'$, $\mathbf{x}_1=\mathbf{x}_2=\ldots=\mathbf{x}_{n'}=\mathbf{0}\in\mathbb{R}^2$, and $\mathbf{x}_{n'+1}, \ldots \mathbf{x}_{2n'}\in\mathbb{R}^2...
2302.13319
https://arxiv.org/abs/2302.13319
2023-02-28 02:16:13
2302
en
32,559
63be946d-d0a7-40ac-9dfe-7e9b9b1ab984
31,100
\section{Introduction} Differential equations are the \textit{de facto} standard for learning dynamics of biological \citep{hirsch2012differential} and physical \citep{tenenbaum1985ordinary} systems. When the observed phenomenon is deterministic, the dynamics are typically expressed in terms of ordinary differential ...
2302.13262
https://arxiv.org/abs/2302.13262
2023-02-28 02:14:33
2302
en
17,454
06608a6e-9c21-488a-9ee1-9b650286d4d1
17,027
\section{Attention Algorithm}\label{sec:low_rank} In this section, we show how to use polynomial approximations for the exponential function in order to approximately perform attention computations. In Section~\ref{sec:low_rank:definition}, we define the type of low-rank matrix approximation which we will use. In Se...
2302.13214
https://arxiv.org/abs/2302.13214
2023-02-28 02:13:10
2302
en
17,683
b1ebdb5c-e28b-460a-9512-6e0b7f8ee54d
17,589
\subsubsection*{\bibname}} \begin{document} \twocolumn[ \runningtitle{Sample Complexity of DR Q-learning} \aistatstitle{A Finite Sample Complexity Bound\\ for Distributionally Robust $Q$-learning} \aistatsauthor{ Shengbo Wang \And Nian Si \And Jose Blanchet \And Zhengyuan Zhou } \aistatsaddress{ Stanford Univer...
2302.13203
https://arxiv.org/abs/2302.13203
2023-02-28 02:12:49
2302
en
13,789
5881f33f-27ed-49a9-b4e5-a247fc9edafa
13,488
\section{Introduction} Deep generative models, such as generative adversarial networks (GANs) \citep{goodfellow2014generative,arjovsky2017wasserstein} and variational autoencoder \citep{kingma2013auto,mohamed2014stochastic}, utilize neural networks to generate new samples which follow the same distribution as the train...
2302.13183
https://arxiv.org/abs/2302.13183
2023-02-28 02:12:01
2302
en
17,322
4ae9b4be-745d-49f2-87ad-a8a80c974565
17,194
\subsubsection*{\bibname}} \bibliographystyle{apalike} \special{papersize = 8.5in, 11in} \setlength{\pdfpageheight}{11in} \setlength{\pdfpagewidth}{8.5in} \newcommand*{\myDots}{\ifmmode\mathellipsis\else.\kern-0.13em.\kern-0.13em.\fi} \newcommand{\mathop{\rm arg~max}\limits}{\mathop{\rm arg~max}\limits} \newcommand...
2302.13093
https://arxiv.org/abs/2302.13093
2023-02-28 02:09:27
2302
en
16,081
382a8acc-acd2-4b02-9621-938f395212ac
15,956
\section{Introduction} The problem of matching two sets of multiple elements, namely set-to-set matching, has received a great deal of attention in recent years~\cite{iwata2015unsupervised,kimura2021shift15m,lisanti2017group,xiao2018group}. The problem is formalized as a task that, given two distinct sets, finds the go...
2302.12991
https://arxiv.org/abs/2302.12991
2023-02-28 02:05:58
2302
en
6,894
d7bb6aef-8086-4156-8b24-9f151eb11cae
6,852
\section{Introduction} This is where the content of your paper goes. \begin{itemize} \item Limit the main text (not counting references and appendices) to 12 PMLR-formatted pages, using this template. Please add any additional appendix to the same file after references - there is no page limit for the appendix. \i...
2302.12909
https://arxiv.org/abs/2302.12909
2023-02-28 02:02:29
2302
en
35,339
058a52ca-1c96-45d1-b3dd-1d352ddeb88c
35,087
\subsubsection*{\bibname}} \bibliographystyle{apalike} \usepackage{bm} \usepackage{graphicx} \usepackage{amsmath} \newcommand{\mathbb R}{\mathbb R} \DeclareMathOperator*{\argmin}{argmin} \DeclareMathOperator*{\argmax}{argmax} \usepackage{amssymb} \usepackage{amsthm} \usepackage{xr} \usepackage{x...
2302.12736
https://arxiv.org/abs/2302.12736
2023-02-27 02:16:19
2302
en
25,715
b2dfdb35-35d1-48bf-9fdc-2ad0b24d27c0
24,982
\subsubsection*{\bibname}} \begin{document} \twocolumn[ \aistatstitle{Understanding the Impact of Competing Events on Heterogeneous Treatment Effect Estimation from Time-to-Event Data} \aistatsauthor{Alicia Curth \And Mihaela van der Schaar} \aistatsaddress{University of Cambridge \And University of Cambridge...
2302.12718
https://arxiv.org/abs/2302.12718
2023-02-27 02:15:52
2302
en
15,568
3c507b94-1d53-4a0c-9d62-6a6bb3239d4a
15,412
\section{Introduction} We consider the following minimization problem: \begin{eqnarray} \theta^*=\argmin_{\theta\in \Theta}\{L(\theta):=\mathbb{E}[l(Z,\theta)]\},\label{eq:model} \end{eqnarray} where $\theta^*\in \Theta \subset \mathbb{R}^d$ is the parameter of interest, $Z$ is a copy of some random vector with an unk...
2302.12717
https://arxiv.org/abs/2302.12717
2023-02-27 02:15:51
2302
en
27,844
f2f224e8-3f50-4e85-b62e-c93ca5e2281f
25,287
\section{Introduction} In the big data era, to deal with data in high dimensional space, a commonly used preprocess tool is dimension reduction. It is well known that Principal Component Analysis (PCA) is one of the most popular techniques for dimension reduction. However, a main drawback of the traditional PCA i...
2302.07712
https://arxiv.org/abs/2302.07712
2023-02-27 02:15:41
2302
en
25,652
eefb999a-d998-4ffc-9cbd-502dc0e1b0ca
25,457
\section{Introduction} \def*}\footnotetext{Equal contribution.}\def\thefootnote{\arabic{footnote}{*}\footnotetext{Equal contribution.}\def*}\footnotetext{Equal contribution.}\def\thefootnote{\arabic{footnote}{\arabic{footnote}} A central question in statistics and machine learning concerns the recovery of useful or int...
2302.12693
https://arxiv.org/abs/2302.12693
2023-02-27 02:15:12
2302
en
13,259
b72b53e4-f86c-4753-bcbe-30dd38ab80dd
13,217
\section{Extension: Multi-stage Pricing} \section{A Detailed Practical Algorithm} In this section, we give a detailed practical implementation of Algorithm \ref{alg: min-max} when the function classes ${\mathcal{H}}$ and ${\mathcal{F}}$ are neural networks (NNs), in which the functions can be parameterized by so...
2302.12670
https://arxiv.org/abs/2302.12670
2023-02-27 02:14:26
2302
en
53,611
9d436e9c-6757-4637-96d2-345db3e99ddc
53,362
\subsubsection*{\bibname}} \setcitestyle{authoryear,round,citesep={;},aysep={,},yysep={;}} \begin{document} \runningauthor{Samuel Holt, Alihan Hüyük, Zhaozhi Qian, Hao Sun, Mihaela van der Schaar} \twocolumn[ \aistatstitle{Neural Laplace Control for Continuous-time Delayed Systems} \aistatsauthor{ Samuel Holt \\ U...
2302.12604
https://arxiv.org/abs/2302.12604
2023-02-27 02:12:24
2302
en
14,374
6dd569db-238e-460b-a335-7301443422cb
14,023
\section{Introduction} Deep neural networks (DNNs) have become the \textit{de facto} solution for a variety of pattern recognition problems due to their ability at modeling deterministic connections \citep{he2016deep}. However, DNNs suffer from many downsides as weak calibration \citep{guo2017calibration} and poor rea...
2302.12565
https://arxiv.org/abs/2302.12565
2023-02-27 02:10:56
2302
en
9,071
b098b0e1-52e6-46bc-8600-6b1476155e8c
8,927
\section{Introduction} Classical results in the area of understanding the expressivity of neural networks are so-called \emph{universal approximation theorems}~\citep{cybenko1989approximation,hornik1991approximation}. They state that shallow neural networks are already capable of approximately representing every con...
2302.12553
https://arxiv.org/abs/2302.12553
2023-02-27 02:10:39
2302
en
10,805
c2100ebb-b4c4-4c6d-a5f5-e7ae330f6730
10,717
\section{Introduction} The minimization of a real-valued function is the most common formulation for mathematical optimization problems. Examples of convex optimization problems in machine learning can be found for instance in \cite{bubeck}. For models involving missing or latent data, \cite{dempster} introduced...
2302.12544
https://arxiv.org/abs/2302.12544
2023-02-27 02:10:31
2302
en
19,138
bf70933d-3c0c-47dd-9c01-a7a84ed161da
19,104
\subsubsection*{\bibname}} \usepackage{silence} \WarningFilter{hyperref}{Token not allowed in a PDF string} \WarningFilter{minitoc(hints)}{W0023} \WarningFilter{minitoc(hints)}{W0024} \WarningFilter{minitoc(hints)}{W0028} \WarningFilter{minitoc(hints)}{W0030} \WarningFilter{minitoc(hints)}{W0039} \usepackage[utf8]{...
2302.12526
https://arxiv.org/abs/2302.12526
2023-02-27 02:09:51
2302
en
26,833
1413b901-243f-41f2-8282-9925822a2223
26,560
\section{Discussion on normality of $W_n$} \label{appendix:gaussianity:Wn} We fix $n,d$ throughout and only consider the case when $K$ grows. In this section, we focus on the distributional limit $W_n \coloneqq \lim_{K \rightarrow \infty} W_n^K$ and discuss whether it is Gaussian. This turns out to be difficult in gen...
2302.05686
https://arxiv.org/abs/2302.05686
2023-02-27 02:09:31
2302
en
165,461
0218aec3-1967-478f-9f8b-70aa2b195bda
165,182
\section{Introduction} Data driven decision making across multiple disciplines including healthcare, epidemiology, econometrics and prognostics often involves establishing efficacy of an intervention when outcomes are measured in terms of the time to an adverse event, such as death, failure or onset of a critical cond...
2302.12504
https://arxiv.org/abs/2302.12504
2023-02-27 02:09:00
2302
en
21,404
59ba1cdd-2d46-4866-8ca5-6f598a40a4d4
21,142
\subsubsection*{\bibname}} \usepackage[utf8]{inputenc} \usepackage[T1]{fontenc} \usepackage{hyperref} \usepackage{xurl} \usepackage{booktabs} \usepackage{amsfonts} \usepackage{nicefrac} \usepackage{microtype} \usepackage{xcolor} \usepackage{graphicx} \usepackage{subc...
2302.12498
https://arxiv.org/abs/2302.12498
2023-02-27 02:08:42
2302
en
20,315
c6a60376-732e-4ad7-b21e-e3733a8401f7
19,971
\section{Introduction} When trained on large-scale natural image datasets, GAN \citep{goodfellow2020generative} is a good approximator of the underlying true image manifold. It captures rich knowledge of natural images and can serve as an image prior. Recently, utilizing the learned prior through GANs shows impressi...
2302.12464
https://arxiv.org/abs/2302.12464
2023-02-27 02:07:27
2302
en
11,573
2df8634d-27f0-416b-a37b-af1ccb389c37
11,049
\section{Introduction}\label{sec:intro} In many real-world reinforcement learning (RL) tasks, limited computing resources make it challenging to apply fully adaptive algorithms that continually update the exploration policy. As a surrogate, it is more cost-effective to collect data in large batches using the current po...
2302.12456
https://arxiv.org/abs/2302.12456
2023-02-27 02:07:13
2302
en
13,312
4056b614-4792-4d99-9b82-0e6a5daa6eca
13,081
\section{Introduction} Pricing American options is a type of optimal control/stopping problem in which the goal is to find the stopping strategy that maximises the option value. Numerically, there have been many attempts based on classical partial differential equation methods and binomial trees \citep{brennan1977valua...
2302.12439
https://arxiv.org/abs/2302.12439
2023-02-27 02:06:42
2302
en
17,531
280da369-d117-4160-9999-b27e0e5f988e
16,829
\subsubsection*{\bibname}} \bibliographystyle{apalike} \usepackage[autonum,colorhypersetup]{shortex} \usepackage{amsfonts} \usepackage{float} \usepackage{graphicx} \graphicspath{ {./figures/} } \usepackage{caption} \usepackage{cleveref} \usepackage{lipsum} \newtheorem{condition}{Condition} \usepackage{xcolor} \use...
2302.12419
https://arxiv.org/abs/2302.12419
2023-02-27 02:05:54
2302
en
27,863
2632ba33-4219-4727-b38c-5a90ccdaf652
27,626
\section{Introduction}\label{sec:introduction} \input{02_introduction} \section{Findings} \subsection{Feature Selection}\label{sec:feature} \input{03a_feature_selection} \subsection{Categorical Handling}\label{sec:categorical} \input{03b_categorical} \subsection{Null Imputations}\label{sec:null} \input{03c_null_impu...
2302.12042
https://arxiv.org/abs/2302.12042
2023-02-27 02:05:16
2302
en
12,499
d7f53036-58b7-43af-a409-ad834ad41e25
12,211
\section{Introduction} \begin{figure} \centering \includegraphics[width=0.99\textwidth]{Figure/images.png} \caption{An SNN graph constructed on top of a $7$-NN graph: here, nodes $i$ and $j$ have $4$ $7$-nearest neighbors; hence, they are SNN neighbors of each other and assigned a ({\it simcos}) edge weigh...
2302.12399
https://arxiv.org/abs/2302.12399
2023-02-27 02:05:08
2302
en
25,093
41471a9e-87a7-4ecb-b909-f7489f5aa7b9
24,925
\section{Introduction} This paper considers linear bandit problems. In this class of problems, a player chooses, in each round $t$, an action $a_t$ from a given \textit{action set} $\mathcal{A}$, which is a subset of a $d$-dimensional linear space. The player then observes the incurred loss $f_t(a_t) \in [-1, 1]$, wher...
2302.12370
https://arxiv.org/abs/2302.12370
2023-02-27 02:03:37
2302
en
15,859
1177278d-bcf7-46eb-a0c6-325c2efc85f2
15,746
\subsection{Binary Classification Setting} \label{sec:binary_classification_loss} We start with the binary classification problem where the labels come from $\{-1,+1\}$. Like in Section 4.1 of~\cite{yin2019rademacher}, we introduce the hypothesis class of linear functions with bounded weights: \begin{equation} ...
2302.12351
https://arxiv.org/abs/2302.12351
2023-02-27 02:02:57
2302
en
80,287
4e16ba17-f333-4aea-b84a-4537dc8d0a93
79,916
\section{Algorithm: Policy Learning via Reward Learning}\label{sec:alg} \begin{algorithm}[t!] \DontPrintSemicolon \KwIn{Number of queries $n$, policy set $C_{\pol}$, oracle $\mathcal{O}_{\rew^*}$} Select $n$ policies $\mathcal{Q} = \{\pi_1, \ldots, \pi_n\}$ and receive noisy reward evaluations $y_i = \mathcal{O}_{\r...
2302.12349
https://arxiv.org/abs/2302.12349
2023-02-27 02:02:49
2302
en
35,655
5eaeb99b-e0aa-44d2-9664-807db641a5ba
35,507
\subsubsection*{\bibname}} \bibliographystyle{apalike} \usepackage{xcolor} \usepackage{amsfonts} \usepackage{amsthm} \usepackage{bbm} \usepackage{float} \usepackage{dsfont} \usepackage{graphicx} \usepackage{hyperref} \hypersetup{ colorlinks, citecolor=black, linkcolor=black, filecolor=black, u...
2302.12341
https://arxiv.org/abs/2302.12341
2023-02-27 02:02:27
2302
en
34,047
edb3b923-d63b-4db3-abba-b64617311829
32,675
\section*{Appendix} \begin{proof}[Proof of Lemma~\ref{lem:intersection_of_bands}.] We can project the density to the span of $u,v$ to get a center-symmetric, isotropic, two-dimensional logconcave density $f:{\mathbb R}^2\rightarrow {\mathbb R}_+$. Let us assume that $u,v$ are orthogonal, the general case when they ar...
2302.12289
https://arxiv.org/abs/2302.12289
2023-02-27 02:00:35
2302
en
26,661
d86fc4dd-e9db-4f82-9056-640069ddbf7f
26,503
\subsubsection*{\bibname}} \usepackage{tablefootnote} \usepackage{tabularx} \usepackage[dvipsnames]{xcolor} \newcommand{\text{\ding{51}}}{\text{\ding{51}}} \newcommand{\text{\ding{55}}}{\text{\ding{55}}} \usepackage{microtype} \usepackage{graphicx} \usepackage{booktabs} \usepackage{algorithm} \usepackage[noend]{algp...
2302.12238
https://arxiv.org/abs/2302.12238
2023-02-24 02:18:23
2302
en
12,507
6e07e3eb-73a0-4e87-a890-2a012446294f
12,323
\section{Introduction} \label{sec: intro} Sampling from a distribution with an unknown normalization constant is a widespread task in several scientific domains. Namely, the goal is to generate samples from a probability measure \begin{align*} \pi(x) \propto e^{-V_*(x)}\,, \end{align*} where $V_* : \mathbb{R}^d \to...
2302.12229
https://arxiv.org/abs/2302.12229
2023-02-24 02:18:11
2302
en
12,186
4ded8657-acb9-4c5f-8eb3-cffb20513c21
12,042
\section{Introduction} Bilevel optimization has drawn significant attention from the machine learning (ML) community due to its wide applications in ML including meta-learning~\citep{finn2017model,rajeswaran2019meta}, automated hyperparameter optimization \citep{franceschi2018bilevel,feurer2019hyperparameter}, reinforc...
2302.04969
https://arxiv.org/abs/2302.04969
2023-02-24 02:17:49
2302
en
15,386
6b181315-3b23-4d96-8d81-4f1b53a67415
15,118
\section{Introduction} The fundamental problem in the theory of Machine Learning is to understand when a given hypothesis class can be learned by an algorithm that has access to finitely many random samples of an unknown objective function. The goal of the learning algorithm, or the \emph{learner} for short, is...
2302.04731
https://arxiv.org/abs/2302.04731
2023-02-24 02:17:43
2302
en
13,367
87b56821-2b22-4f32-b581-9e49986d2fb3
13,279
\section{Introduction}\label{sec:intro} The subject of non-stationary bandit learning has attracted much recent attention. However, what constitutes a non-stationary bandit has yet to be formally defined. Non-stationary bandits have typically been characterized as those for which the ``reward distribution" over time;...
2302.12202
https://arxiv.org/abs/2302.12202
2023-02-24 02:17:28
2302
en
22,706
fc637182-e734-4622-86fe-74915c418d50
22,563
\section{Introduction} \label{sec:int} Multiway data or multidimensional arrays, represented by tensors, are ubiquitous in real-world applications, such as recommendation systems~\cite{frolov2017tensor,zhang2021dynamic}, computer version~\cite{panagakis2021tensor}, medical imaging~\cite{Filippi304} and uncertainty quan...
2302.12148
https://arxiv.org/abs/2302.12148
2023-02-24 02:15:47
2302
en
13,431
0486951a-0088-456e-b613-b12623f4cb92
13,256
\subsubsection*{\bibname}} \input{tables} \input{figures} \input{algorithms} \input{math_commands} \widowpenalty100000 \clubpenalty100000 \begin{document} \runningauthor{Yang Yang, Gennaro Gala, Robert Peharz} \twocolumn[ \aistatstitle{Bayesian Structure Scores for Probabilistic Circuits} \aistatsauthor{Yan...
2302.12130
https://arxiv.org/abs/2302.12130
2023-02-24 02:15:21
2302
en
12,479
c983e4ae-992f-4107-b613-9e7a06dc372c
12,224
\section{Introduction} \subsection{Motivation} This paper is concerned with the issue of detecting changes of a model that lies behind a data stream. {The {\em model} refers to the discrete structural information such as the number of free parameters in the mechanism for generating the data.} We consider the situation ...
2302.12127
https://arxiv.org/abs/2302.12127
2023-02-24 02:15:13
2302
en
30,963
17b59fba-a9bc-4e3f-8b4b-399ab0c2670f
30,283
\section{Introduction} Massive datasets are ubiquitous and require proper storage, computation, and analysis. The amount of available data can be overwhelming and may not fit into a single machine, in which case storing and processing data across multiple machines provides a solution. Furthermore, information is often...
2302.12111
https://arxiv.org/abs/2302.12111
2023-02-24 02:14:56
2302
en
24,908
f3444c3c-5aff-4868-bf90-84ed269835c0
24,692